火力与指挥控制2025，Vol.50Issue(12)：188-197,10.DOI:10.3969/j.issn.1002-0640.2025.12.025
复杂环境下海上小尺度舰船目标识别算法
An Algorithm for Small Scale Ship Recognition in Complex Marine Environments
摘要
Abstract
To address the issues of missed detection and low accuracy for small-scale ships in complex environments,the COS-YOLO algorithm is proposed based on algorithm YOLOv8n.Firstly,the C2f_FasterNext module replaces the C2f module in the backbone network to enhance ship target recognition capability while reducing the numbers of parameters and computation.Secondly,the scale sequence feature fusion module(SSFF)is integrated into the neck network to capture more contextual information,improving small-scale ship feature extraction and reducing missed detection phenomena of small scale ship targets.Then,a dual-branch SMHDet detection head is designed to improve detection accuracy for small-scale ship targets.Finally,the SIOU loss function with a vector angle penalty item is introduced to reduce regression freedom of degree of the model and accuracy and convergence speed of the model is further improved.The experimental results show that compared to YOLOv8n,the mean average precision(mAP)of COS-YOLO increases by 7.9%,the average detection rate increases by 4.9%,and the numbers of parameters and computation drop by 11.3%and 11.0%,respectively.关键词
复杂环境/小尺度/舰船/YOLOV8n/COS-YOLOKey words
complex environments/small-scale/ships/YOLOV8n/COS-YOLO分类
信息技术与安全科学引用本文复制引用
HUANG Hao,SUN Rui,GUO Wei,SHAO Wei..复杂环境下海上小尺度舰船目标识别算法[J].火力与指挥控制,2025,50(12):188-197,10.基金项目
山东省自然科学基金资助项目(ZR2023MF006) （ZR2023MF006）
